Скачать или смотреть What's semi-supervised machine learning? When and how can it be used?

Semi-supervised learning uses a combination of labelled and unlabeled data to train a model. There are many variants for semi-supervised learning (SSL). The oldest and most common variant is called self-learning or self-training or pseudo-labeling. It involves training a small model with limited labelled data, labeling unlabeled data with the model and picking the most confident predictions, training the same or different model with labelled data and predicted data, and repeating the process. Semi supervised learning makes assumptions on the data, including smoothness, low density, and manifold assumptions.

For additional info:
https://www.altexsoft.com/blog/semi-superv...
https://link.springer.com/article/10.1007/...